Added pinecards for the normalising cross section of the CMS 2D mtt yt diff distr.

[enocera]

As the title says.

[cschwan]

Should this be the integrated cross section of CMS_TTB_8TEV_2D_TTM_TRAP? If that's the case I wonder why we don't use that grid and run pineappl sum --integrated CMS_TTB_8TEV_2D_TTM_TRAP.pineappl.lz4 CMS_TTB_8TEV_2D_TTM_TRAP_TOT.pineappl.lz4, or implement that directly into Pineko using a normalization operator that does this. If I remember it correctly we already discussed this, what was our conclusion @felixhekhorn @alecandido?

Summing it with pineappl instead of regenerating the integrated grid has the advantage that the normalizing operation really gives a perfect one as the nominator and denominator really correspond numerically to each. Using separate runs it could be that we get deviations within the MC uncertainty (probably not really a concern in any case).

[felixhekhorn]

if ever such an operation should be implemented inside pinefarm since pineko doesn't know about grids (or it could also go inside vp actually) ... however, I also seem to remember that we already had that discussion and I'm not sure about the outcome either ... but maybe we decided against for some reason ... maybe there was a specific dataset for which this was not true? @scarlehoff do you remember?

actually, looking from the experimental side: what do they do? do they use the full (reconstructed) cross section as denominator?

[cschwan]

actually, looking from the experimental side: what do they do? do they use the full (reconstructed) cross section as denominator?

If we'd like the sum over all bins to be 1 the denominator must be the sum of the numerator bins.

[scarlehoff]

maybe there was a specific dataset for which this was not true? @scarlehoff do you remember?

I think there was some specific dataset for which the integrated cross section was not the integral of the bins (because it was normalized to the inclusive cross section and the bins were not it).

[alecandido]

I think there was some specific dataset for which the integrated cross section was not the integral of the bins (because it was normalized to the inclusive cross section and the bins were not it).

I believe @cschwan or @enocera pointed out, but I don't have any reference...

[alecandido]

Wait, wasn't this the infamous NORM compound?

[enocera]

If I may say two things.

1. Whether the normalisation is such that the integral of the normalised distribution is one or not is experiment dependent. The way to check it is to take the experimental values and see if they integrate to one. This is what happens for the specific measurement in this PR.
2. The reason why I didn't use the option pineappl sum --integrated CMS_TTB_8TEV_2D_TTM_TRAP.pineappl.lz4 CMS_TTB_8TEV_2D_TTM_TRAP_TOT.pineappl.lz4 is that I was unaware of it. I've checked that the ratio between the numerator integrated over the bin widths and the denominator was giving one (it was giving 1.0001). I agree with @cschwan that using pineappl sum instead of producing independent grids is better (and quicker). So @cshwan, let me know if you think that this PR should be closed, and if I have to replace the grid in Added pineapplgrid for CMS_TTB_8TEV_2D_TTM_TRAP normalising cross section pineapplgrids#28 with that produced with pineappl sum (and regenerate the FK table)
   All this is of course more a matter of tidiness than a matter of substantial changes in the results.

[cschwan]

Given that the cut is potentially wrong (see my review above) I'd prefer to pineappl sum it.

[scarlehoff]

Shouldn't we merge this?

[enocera]

@scarlehoff My understanding is that this is not needed in principle, because one can compute the normalising grid just with pineappl sum --integrated. So I'm inclined to say that this PR is closed w/o merging.

[scarlehoff]

But since the fktable is used during the fit, shouldn't we have then the --sum option done automatically (maybe by the card for CMS_TTB_8TEV_2D_TTM_TRAP) otherwise we won't be able to regenerate the grid.

/cc @alecandido @felixhekhorn

[enocera]

Ah, OK. Then let's do things easy this time. Let's merge this.

[felixhekhorn]

Ah, OK. Then let's do things easy this time. Let's merge this.

Agreed - let's do the easy stuff first.

Later this might be a varient of NNPDF/pinefarm#3 where here "external" means "from another pinecard"